diin papr riftp.iza.org/dp10710.pdf · 2017-04-25 · ara iza dp no. 10710 april 2017 information,...

DISCUSSION PAPER SERIES

IZA DP No. 10710

Leonardo BecchettiVittorio PelligraTommaso Reggiani

Information, Belief Elicitation and Threshold Effects in the 5X1000 Tax Scheme: A Framed Field Experiment

APRIL 2017

Any opinions expressed in this paper are those of the author(s) and not those of IZA. Research published in this series may include views on policy, but IZA takes no institutional policy positions. The IZA research network is committed to the IZA Guiding Principles of Research Integrity.The IZA Institute of Labor Economics is an independent economic research institute that conducts research in labor economics and offers evidence-based policy advice on labor market issues. Supported by the Deutsche Post Foundation, IZA runs the world’s largest network of economists, whose research aims to provide answers to the global labor market challenges of our time. Our key objective is to build bridges between academic research, policymakers and society.IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author.

Schaumburg-Lippe-Straße 5–953113 Bonn, Germany

Phone: +49-228-3894-0Email: [email protected] www.iza.org

IZA – Institute of Labor Economics

DISCUSSION PAPER SERIES

IZA DP No. 10710

Information, Belief Elicitation and Threshold Effects in the 5X1000 Tax Scheme: A Framed Field Experiment

APRIL 2017

Leonardo BecchettiUniversity of Rome Tor Vergata and CESIS

Vittorio PelligraUniversity of Cagliari and CRENoS

Tommaso ReggianiLUMSA University Rome and IZA

ABSTRACT

APRIL 2017IZA DP No. 10710

Information, Belief Elicitation and Threshold Effects in the 5X1000 Tax Scheme: A Framed Field Experiment*

In this paper we study by means of a framed field experiment on a representative sample

of the population the effect on people’s charitable giving of three, substantial and

procedural, elements: information provision, belief elicitation and threshold on distribution.

We frame this investigation within the 5X1000 tax scheme, a mechanism through which

Italian taxpayers may choose to give a small proportion (0.5%) of their income tax to a

voluntary organization to fund its activities. We find two main results: a social information

effect, since information on total donations received by the organizations in the previous

year significantly increases the share of donors, and a distributional effect, leading, the

information provision, to a significant increase in the share of donors to the organization

reporting the lowest aggregate donations.

JEL Classification: C91, D64, H00

Keywords: charitable-giving, framed field experiment, social information effect, 5X1000

Corresponding author:Vittorio PelligraDepartment of Economics and BusinessUniversity of CagliariV. le S. Ignazio, 1709123 CagliariItaly

E-mail: [email protected]

* Financial support from CSV-Sardegna Solidale is gratefully acknowledged.

2

1. Introduction

In times of tightened government budget constraints in high-income countries the subsidiary

role of not-for-profit organizations in the provision of public goods and services is becoming

of paramount importance. The supply of such goods and services crucially depends on the

organization’s capacity of raising funds and stimulate individuals’ willingness to give. It

therefore becomes increasingly relevant to understand better how different funding schemes

may work and which factors may affect people’s willingness to give. Economists have

traditionally focused their attention on the effect of changes in ‘price’ on giving responses. To

identify the effect of ‘price’ changes on donations and the associated elasticity of giving is

especially important from the policy makers’ perspective. Since, in fact, charitable giving

favors the production of public goods that the public sector might want to encourage, donations

can subsidize it by, for instance, tax deductibility, as it happens in many fiscal systems around

the world. In the U.S., for this reason, the price of charitable giving is inversely related to the

marginal tax rate; that implies that those with higher incomes get higher marginal subsidies. In

the U.K. a similar logic applies to the payroll giving (Give as You Earn). In this system

donations are removed before income tax is calculated and deducted. The consequence is that

tax is calculated on a lower amount and the deduction may lower the amount of tax to pay also

by changing individual’s tax bracket.

There are fiscal systems even more generous, where the tax deductibility appears in extreme

forms. The Italian 5X1000 (five by thousand), introduced in 2006, for instance, permits the

taxpayers to donate each year to their preferred organization a share equal to the five per

thousand of their personal income tax. In this way, the State finances the not-for-profit sector

forgoing a certain amount of tax revenues and delegating to the taxpayer the control on how to

3

allocate this public fund among the different organizations. This represents an extreme form of

‘price’ reduction with the actual price set equal to 0.

The efficacy of these systems in stimulating private funding depends on how tax-payers and

other potential donors respond to the formal, informal and social incentives they provide. In

this respect, one such element that both economists and psychologists have recently focused on

is social information: how information about other people’ contributions, affect individuals’

willingness to fund charities and public goods more in general1. In general, we may think of

two types of influences of social information: a “substitution effect” and a “complementarity

effect”. The former is consistent with models of altruistic behavior (Becker, 1974; Warr, 1982;

Roberts, 1984; Andreoni, 1988 and 1990): the more other people (or institutions) give, the less

the individual is motivated to contribute. The second effect is implied by models of reciprocity

(Sugden, 1984), conformity (Bernheim, 1994), signalling (Vesterlund, 2003) social image

(Benabou and Tirole, 2006; Andreoni and Bernheim, 2007; Ellingsen and Johannesson, 2007,

2008), and social connection (Scharf, 2014). They all predict that knowing people are giving

stimulates other people’s donations.

Empirical evidence has been, so far, largely inconclusive, providing support to both effects:

Andreoni (2006), Powell and Steinberg (2006) and Kolm and Ythier (2006) review a number

of empirical studies concluding that, apart from complete crowding-out, none of the other

predictions (complementarity or substitutability) can be disconfirmed. Data from lab

experiments (see Shang and Croson, 2009 for a survey) are generally supportive of the positive

effect of social information but the robustness of these results has been variously criticized,

mainly because of weaknesses in the experimental designs and on the questionable external

1 See Andreoni (2006) and Vesterlund (2006) for economics-oriented reviews and Cialdini and Goldstein (2004),

Penner et al. (2005), Weber et al. (2004) for reviews about contributions from psychology.

4

validity of their conclusions. Few studies have tried to overcome such criticisms by adopting

more convincing treatment manipulations and considering non-conventional subject pools. In

one of these, Frey and Meier (2004) conduct a natural field experiment with all the students of

the University of Zurich. They collect data supporting the complementarity prediction, as they

find that, on average, people contribute more often to a charity if they know that many others

contribute. While this study is focused on the willingness to contribute or not, to a public good,

Shang and Croson (2009) examine the effect that information about the amount of others’

donation may have on the amount of participants’ contribution. This study is fielded in the

context of a fund-raising campaign for a public radio and finds a positive and significant relation

between social information and the amount given.

We contribute to this growing literature by focusing on the effect of three, substantial and

procedural, elements: social information, belief elicitation and the imposition of an exogenous

threshold on total contribution. We study both the impact of these elements on individuals’

willingness to donate and, more importantly, how they affect the distribution of donations

among a set of different organizations. We investigate these elements by means of a framed

field experiment involving a representative pool of the population. We frame the experiment

considering the ‘5X1000 tax scheme’, a mechanism through which Italian taxpayers may

choose to give a small proportion of their income tax to a voluntary organization to fund its

activities. The design and the hypotheses of our experiment are motivated, as we said, by an

interest for the role of information in charitable giving and its policy implications, but also by

the heated debate among policy-makers and practitioners fueled by different positions about

the role of public-social information, which is accompanying in these years, the

experimentation of the 5X1000 mechanism and the discussion about the pros and cons of its

implementation.

5

In its basic form this mechanism allows taxpayers to divert, each year, a small share (equal to

the five per thousand) of their personal income tax from the State to one among not-for-profit

organizations active in the production of public goods2. Through this channel, the State finances

the no-profit sector forgoing a certain amount of tax revenues and delegating to the taxpayer

the control on how to allocate this portion of public funds among the different organizations

and their preferential ends.

In the recent years, however, the implementation of the 5X1000 mechanism has been criticized

on two main points: first, the neglect of the role of information about each organization’s

funding results. Data about how much each of them received in the previous years, in fact, has

not been made available on a regular basis. Many argue that knowing this data is crucial to a

well-informed choice and that they may have important consequences on the allocation of

funds. First, they may impact on the decision to donate. Secondly, if we assume that people not

only care about the total amount distributed to the organizations but also about how these funds

are allocated among the organizations, knowledge about the previous allocations of funds may

lead taxpayers to choose to redistribute from the top organizations to the bottom, if they like

pluralism and diversity or they aim at maximizing the impact of their single donations. On the

contrary, one may hypothesize that the same information, may, on the contrary, induce people

to give more to the top organizations. In fact, if donors interpret giving from other taxpayers as

signal of quality, for instance, they may want to reward and make even stronger the best

performing organizations.

2 More specifically are eligible to receive donations all the voluntary organizations and other foundations and

organizations (public or private) operating in the fields of scientific research, higher education, public health

(including non-professional sport associations), cultural promotion and environmental protection, that ask to be

included in an official list maintained by the tax authority.

6

The second reason of debate refers to the decision by the Ministry of Finance to impose an

upper limit to the total amount to be actually distributed. In other words, each year the Ministry

of Finance, establish the maximum amount to be distributed, irrespectively of the actual amount

gathered from taxpayers’ choices. If total donations exceed the threshold set for a given year,

only a fraction of each donation will be distributed, to a sum equal to the fixed threshold.

A third issue that we focus on refers to the role of donors’ prior beliefs. Social psychologists

documented that being asked to explicitly stating one’s beliefs may, in different context, affect

subject’s choices who, in an attempt to achieve consistency may modify their intended behavior

(Cialdini, 1984). Economists have both theoretically and experimentally studied this ‘taste for

consistency’ and found that is a powerful determinant of behavior (Yariv, 2005; Johansson-

Stenman and Svedsater 2008; Falk and Zimmerman, 2013). We study this ‘belief elicitation

effect’ and its consequences on giving. More specifically we test whether simply asking to

formulate beliefs about how each organization ranked in terms of aggregate donations produces

significant changes in donors’ behavior. We also estimate whether participants’ beliefs are

systematically biased. Such bias in fact, may lead to a change in giving behavior when

information about the real figures are provided.

We find six main results with respect both to the probability to give and to the donations’

distribution. First, when information about the previous year’s aggregate donations received by

each organization is provided, the probability to donate becomes significantly higher (from 90

to 96 percent) with respect to when information is not available. Second, we find that the belief

elicitation procedure (making the ranking salient) increases significantly the probability to

donate as well (from 90 to 97 percent). Third, the imposition of the threshold to the amount to

be distributed decreases significantly the relative probability to donate of about 5 percentage

7

points. The fourth finding refers to the distributive effect of the information provision. When

the amount of funds received by each organization in the previous year is made known, we

observe a statistically significant redistribution from the top organizations toward those at the

bottom. Fifth, although participants’ beliefs are systematically inaccurate, the elicitation

process leads to a statistically significant increase in the contribution to the best performing

organizations. Finally, we find that the imposition of the threshold up to the beliefs elicitation

does not have any significant redistributive effects.

The remainder of the paper proceeds as follows. The next section describes the experimental

design, the hypotheses and the procedures. Section 3 summarizes our findings. Section 4

concludes.

2. The Experiment

2.1. Design

Our framed field experiment is designed to investigate the effect of information disclosure,

beliefs elicitation and the imposition of an exogenous threshold on charitable giving in a large

and representative sample of the population. The baseline experimental task is a modified

version of the ‘generosity game’ (Güth, 2010; Güth et al., 2012) in which all the experimental

subjects play the role of the dictator, while the recipient is a real not-for-profit organization.

Each dictator receives a lottery ticket and has to decide whether and to whom to donate (by

selecting one organization from a list of nine provided by the experimenters3) the monetary

3 The list of organizations eligible for the 5X1000 contribution includes about 50 thousand entries (year 2016). We

considered a selection of the most well-known among those that operate nationwide. We tried to diversify both in

terms of total amount received and mission of the organization. We included, in alphabetical order: ADMO – Bone

Marrow Donors Association, Amnesty International (Italian Section), Caritas, Emergency, Fondazione Banco

8

prize (1000 euros) in the event of winning the lottery (see section 2.3 for more details). The

dictator can as well decide not to donate but, in that case, the share of the pie going to the

dictator is set equal to 0. We consider this game instead of the more common dictator game, in

order to mimic as close as possible the real decision faced by the taxpayers in the ‘5X1000’

mechanism. Not giving, in fact, does not increases the taxpayer wealth. If they opt not to give

through the 5X1000 mechanism, the entire amount of their taxes will be retained by the State.

In standard lab experiments participants usually receive a monetary endowment at the

beginning of the experiment like “manna from heavens”; this introduces an element of

unrealism that has been repeatedly criticized. In our design, on the contrary, reality and the

simulated situation are much more closely related since both taxpayers and participants decide

to donate “out of pocket” money that in any case would not go back to them in case of decision

not to donate. This close resemblance between experimental procedure and real-life choice

increases the external validity of our design.

The game is played in a between-subject design, in four different treatments (Table 1). In the

first, “No Information Treatment” (NoInfoT), subjects play the baseline game with no additional

information (see Appendix A for the instructions). In the second, “Information Treatment”

(InfoT), players are given information about the donations received by each of the nine

organizations in the previous fiscal year through the 5X1000 mechanism. In the third, “No

Information plus Belief Elicitation Treatment” (BeliefT), players have no information but they

are asked to rank the organizations according to their beliefs about the amount of donations

received in the previous year through the 5X1000 mechanism. Beliefs are elicited at the same

moment the donation choice is done as it is clear from the attached questionnaire where beliefs

Alimentare Onlus, Greenpeace, L.A.V. Lega Anti-vivisezione, UNICEF (Italian Section) and WWF - World

Wildlife Foundation Italy. For a detailed description of the organizations see the Online Supplementary Material.

9

and donations choices appear as right and left columns of the same line (Appendix A). Beliefs

are not incentivized. In the fourth, “No Information plus Belief Elicitation and Threshold

Treatment” (Belief&ThresholdT), players are not given information, are asked their beliefs and

their choices are subject to a threshold. More specifically participants are told that, if the

aggregate amount of contributions donated to all the organizations passes a maximum,

unspecified, amount, each organization will receive only a percent of the amount actually

donated to it. This feature of our design is aimed at reproducing the actual threshold mechanism

existing in the Italian 5X1000 donation system where, if total donations exceed the threshold

imposed each year by the Ministry of Finance, only a proportional fraction of each donation is

actually distributed, up to a sum equal to the maximum fixed amount. In the experiment the

threshold is unspecified because in the real system, the total amount to be distributed each year

is decided by the Ministry of Finance ex-post, after taxpayer have already taken the decision

whether to give or not.

TABLE 1 ABOUT HERE

We implemented an incentive system similar to that used by Fong and Luttmer (2011) and

Pelligra and Stanca (2013) which is particularly suited for experiments with large sample of

subjects. We gave to each respondent a lottery ticket. Once all the choices were collected, one

of the tickets was randomly drawn and the choices that the owner of that ticket made determined

both the actual payments (1000 euros or nothing) and eventually the recipient (the selected

organization). The participant, as well as the organization, were then contacted by SmartLab,

10

the market research firm that fielded the research4, and informed about the result of the lottery

and, in the case of the organization, paid by bank transfer. We ensured anonymity using a

system of unique codes to identify the donors whose identity remained unknown to the

experimenters. The participants do not know the total number of the interviewees, therefore

they cannot infer the probability of being selected to be paid for real. The different stages of the

lottery were filmed and made available to all participants upon request. In the context of our

experiment, this two-stage procedure was easier to implement than the usual one-stage random

lottery incentive system, while ensuring, we think, the necessary salience.

2.2. Hypotheses

From our design we can infer a series of testable hypotheses. First, let S(g)T denote the share of

players who give, and S(ng)T the share of players who decide not to give in treatment T =

{InfoT, NoInfoT, BeliefT, Belief&ThresholdT}; similarly let GjT the total amount of donations

received by the organization j in treatment T.

The hypotheses we test refer to the effect of the information disclosure, the beliefs elicitation

and the introduction of the threshold both on the probability of giving (hypotheses a) and on

the distribution of the donations (hypotheses b).

Hypothesis 1a

(information effect on probability of giving):

H0: S(NoInfoT) = S(InfoT)

HA: S(NoInfoT) ≠ S(InfoT)

Hypothesis 1b

(information effect on the distribution):

H0: Gj(NoInfoT) = Gj(InfoT)

HA: Gj(NoInfoT) ≠ Gj(InfoT)

4 www.smartlabkaralis.it. The research was funded by CSV-Sardegna Solidale (www.sardegnasolidale.it).

11

The goal of these hypotheses is to investigate the effect on charitable behavior of the

information about the total amount of funds raised by each organization in the previous year.

To this end we compare the share of givers and their preferred organizations in the

“Information” and in the “No information” treatments. Under the null hypothesis the

information has no effects.

Hypothesis 2a

(belief effect on probability of giving):

H0: S(NoInfoT) = Sj (BeliefT)

HA: S(NoInfoT) ≠ Sj (BeliefT)

Hypothesis 2b

(belief effect on the distribution):

H0: Gj(NoInfoT) = Gj (BeliefT)

HA: Gj(NoInfoT) ≠ Gj (BeliefT)

The goal of these hypotheses is to test the effect on charitable behavior and its distribution of

the elicitation of prior beliefs about the donations received. Under the null hypothesis prior

beliefs have no effects.

Hypothesis 3a

(threshold effect on probability of giving):

H0: S (BeliefT) = Sj (Belief&ThresholdT)

HA: S (BeliefT) ≠ Sj (Belief&ThresholdT)

Hypothesis 3b

(threshold effect on the distribution):

H0: Gj (BeliefT)= Gj (Belief&ThresholdT)

HA: Gj (BeliefT) ≠ Gj (Belief&ThresholdT)

With this hypothesis we test whether imposing a threshold on the actual distribution of the total

donations given by the contributors affects the probability to donate. Under the null hypothesis

the threshold has no effects.

12

2.3. Procedures

The experiment was conducted in August-September 2013 using a representative sample of

2000 adults stratified by gender, age and education. Subjects responded to a questionnaire

carried out in Sardinia, an autonomous region of Italy, using the PAPI (Paper-and-Pencil

Interviewing) technique. In addition to a series of questions on socio-demographic

characteristics, values, beliefs, and pro-social activities, we included the choice task described

in the previous section5.

Participants are contacted by the interviewers at home and invited to participate in a simple

experimental study conducted by the University of Cagliari. They receive a self-explaining

anonymous questionnaire that is completed autonomously without any intervention from the

interviewer. After answering questions about socio-demographic characteristics, each

participant is informed about the rules of the choice task, the consequences of her choices and

the functioning of incentive system. They are then presented with the task and their

understanding of the game is checked by a series of control questions. After completing the

choice task, they answer the remaining sections of the survey. Completing the entire

questionnaire takes on average about 20 minutes.

The main benefits of using a representative sample of the population (see Table 2 for the sample

structure), instead of the usual convenience pool of student subjects, are relative to the sample

size and to the heterogeneity and representativeness of subjects whose choices are observed.

These factors have obvious pros in terms of a finer understanding of the mechanism underlying

decision-making, since we have a larger number of socio-demographic determinants to which

5 The instructions and are provided in the Appendix and the questionnaire is available in the Online Supplementary

Materials.

13

we can relate variations in behavior, but also in terms of external validity and generalizability

of the experimental results.

TABLE 2 ABOUT HERE

Once all the questionnaires were collected we run the lottery that determined the organization

to be paid. At this stage both the chosen participant and the organization were informed of the

result.

3. Results.

We contacted 2000 subjects and about 75% of them consented to participate in the study.

Overall, we obtained 1405 complete questionnaires, 214 questionnaires have been dropped

because in the two treatments featured by beliefs elicitation, beliefs were incomplete or

inconsistent. This brings the sample to 1191 observations (Table 3).

TABLE 3 ABOUT HERE

Descriptive findings document that our sample is almost perfectly balanced in terms of gender

(53 percent females), with average age being 45. Respondents live in households with 2.2

components on average and around 46 percent of them are married. Non-donors6 are 6 percent

in the overall sample. This share remains very low also if we look at it in the four different

6 We call for simplicity non-donors those who choose not to donate to any of the listed organizations.

14

treatments (highest at 9.7 percent in the NoInfoT and lowest at 2.8 percent in the BeliefT) (see

Table 4).

TABLE 4 ABOUT HERE

3.1. Treatments effect on the probability to donate.

Figure 1 depict the empirical distribution of donation across the nine organizations. The first

outcome of interest is represented by the effect of the different treatments on the probability to

generate an actual donation (hypotheses 1a, 2a and 3a).

Table 5 (model 1)7 displays estimations based on an intuitive OLS Linear-Probability-Model

in which the outcome dummy variable Donation (1= donation, 0=non-donation) is regressed on

treatment dummies InfoT, BeliefT, Belief&ThresholdT. The baseline treatment NoInfoT is

omitted for collinearity reasons and it is captured in the constant term. The average donation

rate under the baseline treatment is 90%. InfoT has a positive and significant effect on the

probability to donate increasing it by 7 percentage points (in relative terms, the propensity to

donate increases by 8%). BeliefT has a positive and significant effect on the probability to make

the donation increasing it by 6 percentage points (in relative terms, the pool of donors increased

by 7%). Belief&ThresholdT combines together beliefs elicitation (Belief) and a threshold rule

(Threshold). The coefficient for this treatment is very small and not statistically significant at

any conventional level. This means that the introduction of a threshold rule brings a relative

7 Model 2 in Table 1 replicates the very same findings of Model 1, controlling for further demographics such as

age, gender and education level.

15

significant, negative and sizable (-5 percentage points) effect in terms of propensity to donate

since it neutralizes the positive and significant effect generated by BeliefT.

TABLE 5 AND FIGURE 1 ABOUT HERE

3.2. Information effect (NoInfot vs InfoT)

Providing information about past donations leads to a polarization of the donations in favor of

(2 out of 3) the less endowed charities (Table 6 and Figure 2). The qualitative polarization of

the categorical distribution of recipients, is statistically significant according to Person �� test

(p-value<0.001 | tag: a). 8 The non-parametric test rejects the null hypothesis of independence

between experimental conditions and allocation of the donations to the different recipients.


This result is even more clear-cut if the different charities are clustered according to the volume

of funding actually rose during the previous 5x1000 wave (Table 14 and Figure 1). Cluster A

groups together the two major charities / outliers (Emergency and Unicef) that are able to attract

a great volume of donations ( >5 millions euros, each). Cluster B hosts organizations able to

raise about € 1 million each (LAV, WWF Greenpeace, Amnesty). Cluster C represents the three

8 All significant (p-value<0.01) non-parametric tests are jointly significant at the 5% level according based on

Bonferroni–Holm stepwise multiple-hypotheses-testing procedure (Holm 1979). See Table 13 for the criterion cut-

off p-values.

16

less endowed organizations (Caritas, Banco Alimentare, ADMO) collecting less than 200,000

euros each. Providing information leads donors to act systematically more in favor (with respect

to the no information treatment) of the less endowed charities of cluster C.

The previous result is confirmed also in this alternative clustered setting (see Table 7 and Figure

3). The categorical distribution of recipients under InfoT is contrasted against the NoInfoT

distribution. Donations density moves from cluster A (-5 percentage points) and B (-11

percentage points) in favor of cluster C’s charities (+ 16 percentage points). This qualitative

polarization of the donation behavior results to be highly statistically significant according to

Person �� test (p-value<0.001 | tag: b).


3.3. The beliefs effect (NoInfoT vs BeliefT)

In the BeliefT treatment we ask participant to rank the organizations in terms of how much they

believe each organization received the previous year. We do this to verify the effect of priming

a comparative decision rule in the participants that could lead them, not only to focus on their

preferred organization but also to think in terms of global distribution of funds. Besides we are

also interested in studying whether beliefs are generally aligned with actual donations or

systematically biased. This latter case, in fact, could explain any changes due to the provision

of information.

We find that the belief elicitation procedure leads to a bipolarization of the donations with

respect to the NoInfoT baseline case (Table 8).

The bipolar U-shaped categorical distribution of recipients resulted by the introduction of the

beliefs condition is statistically different from the distribution generated under the NoInfoT

according to Person �� test (p-value=0.003 | tag: c). The non-parametric test rejects the null

17

hypothesis of independence between experimental conditions and donation allocations to the

different recipients.


Also in this case, the analysis at cluster level helps to better highlight the underling dynamic

(Table 9 and Figure 5). Visual inspection of Figure 5 suggests how the U-shaping effect is

mainly driven by an increase of donations in favor of better endowed charities belonging to

cluster A (+ 10 percentage points) and a correspondent decrease of donations to the less

endowed ones of cluster B (-8 percentage points) and C (-2 percentage points). As for the

previous test, also in this further clustered configuration the Person �� test rejects (p-

value=0.008 | tag: d) the null hypothesis of independence between experimental conditions and

donation allocations to the clustered recipients.


3.3.1 Accuracy of beliefs and “beliefs-donation” correlation.

The degree of accuracy of the beliefs is very low: only the 0.5% of the pool of subjects reported

a ranking in line with the actual one; 2.4% of the subjects ranked correctly the top three

organizations; the 2.6% ranked correctly the three organization in the middle range of the list

(ranks 4, 5 and 6); the 4.7% correctly identified three organizations at the bottom of the list.

Subjects tend to donate to the organization they rank first. This correlation (Table 10) is robust

and statistically highly significant (Spearman correlation: +0.39, p-value<0.001).

18

TABLE 10 ABOUT HERE

Symmetrically, subjects tend to systematically avoid donations in favor of the charities they

assign the lowest rank (Spearman correlation: -0.18, p-value<0.001). This “beliefs effect” goes

in the opposite direction with respect to the “information effect”. While under information

subjects tended to favor the less endowed organizations, under the beliefs treatment subjects

granted generosity to the organizations that in their opinion are able to attract more donations.

3.4. Threshold effect (BeliefT vs Belief&ThresholdT).

From the distributional perspective, the effect of the threshold rule does not imply major effects.

The categorical distribution of donations generated by Belief&ThresholdT (Table 11 and Figure

6) is statistically indistinguishable compared the one realized under BeliefT according to Person

�� test (p-value>0.3 | tag: e).


The same result holds true when the clustered categorization is considered. Visual inspection

of Figure 7 (see also Table 12) confirms how the two clustered categorical distributions display

analogous patterns of donations by treatments (cluster A≈35%; cluster B≈16.5%; cluster

C≈48.5%). The Person �� test fails to reject (p-value>0.7 | tag: f) the null hypothesis of

independence between experimental conditions and donation allocations to the clustered

recipients.

19


4. Conclusions.

In this paper, we report on a framed field experiment designed to study the effect, on people’s

charitable giving, of three, substantial and procedural, elements: information provision, belief

elicitation and threshold on distribution. The study is framed within the 5X1000 tax scheme, a

mechanism actually available to Italian taxpayers, which allows them to allocate a small part

of their taxes to a selected not-for-profit organization. The Italian law allows only a single

choice option since it is not possible to allocate the amount to more than one organization. We

replicate the scheme by asking to a representative sample of the population to allocate a

monetary prize to one among a list of well-known not-for-profit organizations. In order to

mimic the tax donation scheme where there is no gain for the taxpayer in case of no choice, we

consider a variant of the generosity game where the prize is not earned by the dictator in case

she chooses not to donate. Based on these characteristics, our experimental design allows us to

test three treatment variables associated with three different potential effects by: i) providing

information on the aggregate amount of donations received through the 5X1000 channel by the

listed organizations in the past year; ii) eliciting participants’ beliefs on the organization rank

in terms of donations received in the past year; iii) creating an upward bound on aggregate

donations which imposes a proportional cut to contributions to each organization in case the

aggregate threshold is passed (as it is in the actual 5X1000 mechanism).

We find two main sets of results: first on how information, beliefs elicitation and threshold

affect the likelihood of giving. In this respect, the information effect on the willingness to

donate is consistent with what we define the “complementarity effect”. Knowing, in fact, how

much others have donated in the recent past, the probability to donate increases significantly (7

20

percentage points). A similar effect is observed when, before choosing, players are asked to

manifest their beliefs about how each organization rank in terms of previous donations. Priming

this comparative frame drives people to give more (6 percentage points). A negative effect is,

instead, associated to the imposition of a threshold to the maximum amount to be distributed.

The threshold effect reduces relative participants’ willingness to give of about 5 percentage

points.

The second class of findings refer to the distributive effects. What is the impact of the treatment

variables on the distribution of the donations? These are, we think, our main and more novel

findings. If we consider donations that can alternatively target different organizations we see

that both information provision and belief elicitation exert a significant distributive effect: with

respect with the baseline no-information treatment, when information is provided, participants

tend to significantly lower their contributions to the top organizations and to be more generous

with those at the bottom of the ranking. On the contrary, when beliefs are made salient the

elicitation process leads to a statistically significant modification in the distribution. When

asked to explicitly rank the organizations according to their beliefs about previous funding

participants give systematically to the organization they think is the best performer. Globally,

this process leads to a distribution where the top performing organizations receive more and the

worst performing organization slightly less. Finally, we find that the imposition of the threshold

does not have any significant redistributive effects.

Objective information, seems to work against people’s expressed beliefs. In the first case, in

fact, participant prefer to donate to the poorer organizations: it could be to maximize the

marginal impact of their money or to foster pluralism and finance different forms of

intervention. Disentangling the reasons behind such a behavior will be a subject for further

research. We know, however, that when individual’s beliefs are made salient, these lead

21

participants to concentrate their donation towards the best performers (with only two

exceptions, Caritas and ADMO). We may speculate that the belief-elicitation process could

prime in the participants a comparative / competitive frame that tends to induce the participant

to donate more frequently to the organization they rank first, producing, in this way, a sort of

“St. Matthew effect”9 (Merton, 1968; Salganik et al., 2006) which consists in giving more to

those who have already received a lot and less to the poorer organizations. The same process

may be at work also in NoInfoT. People give more to the organizations that received less, but

since beliefs are systematically inaccurate, the organizations that receive more in NoInfoT may

well be those which ranked first according to participants’ belief. With our data, we cannot

control for that because we elicited beliefs in a no-information setting. However, this argument

stresses the fact that the two seemingly inconsistent results, giving less with information and

more to the top organizations with belief-elicitation, are not necessarily against each other.

In concluding we think that these findings, especially those related to the distributional effect,

are novel and may have important policy implications, both in general and more specifically in

the context of the 5X1000 mechanism. The introduction of a threshold for the total amount to

be distributed, for example, if on the one hand reduces the State’s expenditure, at the same time,

could crowd-out private donations, leading to an under-financing of the no-profit sector.

Secondly, if one, instead, wants to encourage donations and foster pluralism, the social

information effect could be an effective tool of behavior change, both at the public level and

also for each organization. Especially the poorer ones may want to communicate to the

prospective donors the funds collected by their richer competitors in the past. This could, in

fact, favor redistribution to their advantage.

9 “For unto every one that [we believe] hath shall be given, and he shall have abundance: but from him that [we

believe] hath not shall be taken even that which he hath” (Matthew 25:29, King James Version).

22

References

Andreoni, J., 1988. Privately provided public goods in a large economy: the limits of

altruism. Journal of Public Economics 35 (1), pp. 57–73.

Andreoni, J., 1990. Impure altruism and donations to public goods: a theory of warm

glow giving. Economic Journal 100 (401), pp. 464– 477.

Andreoni, J., 2006. Philanthropy. In Kolm, S.-C., Mercier Ythier, J. (Eds.), Handbook of

Giving, Reciprocity and Altruism (pp. 1201–1269). Amsterdam: North Holland.

Andreoni, J., Bernheim, D., 2009. Social image and the 50–50 norm: A theoretical and

experimental analysis of audience effects. Econometrica 77(5), pp. 1607-1636.

Becker, G.S., 1974. A theory of social interactions. Journal of Political Economy 82, pp.

1064- 93.

Benabou, R., Tirole, J., 2006. Incentives and Prosocial Behavior. American Economic

Review 96(5), pp. 1652– 1678.

Bernheim, D., 1994. A Theory of Conformity. Journal of Political Economy 102(5), pp.

841-877.

Cialdini, R., 1984. Influence, the Psychology of Persuasion.New York: Harper Collins.

Cialdini, R.B., Goldstein, N.J., 2004. Social influence: Compliance and conformity.

Annual Review of Psychology 55, pp. 591–622.

Ellingsen, T., Johannesson, M., 2007. Paying Respect. Journal of Economic Perspectives

21(4), pp. 135–49.

Ellingsen, T., Johannesson, M., 2008. Pride and Prejudice: The Human Side of Incentive

Theory. American Economic Review 98(3), pp. 990-1008.

Falk, A., Zimmermann, F., 2013. A Taste for Consistency and Survey Response

Behavior. CESifo Economic Studies 59(1), pp. 181-193.

23

Fong C., Luttmer, E., 2011. Do Race and Fairness Matter in Generosity? Evidence from a

Nationally Representative Charity Experiment. Journal of Public Economics 95(5 - 6), pp.

372 – 394.

Frey, B. S., Meier, S., 2004. Pro-social behavior in a natural setting. Journal of Economic

Behavior & Organization 54, pp. 65–88

Güth, W., 2010. The Generosity Game and calibration of inequity aversion. Journal of

Socio-Economics 39, pp. 155–157.

Güth, W., Levati, M.V., Ploner, M., 2012. An experimental study of the generosity game.

Theory and Decision 72(1), pp 51-63.

Holm, S., 1979. A simple sequentially rejective multiple test procedure. Scandinavian

Journal of Statistics, 65–70.

Kolm, S.C., Ythier, J.M., 2006. Handbook of Giving and Reciprocity and Altruism,

Amsterdam: North Holland.

Merton, R. K., 1968. The Matthew Effect in Science. Science. 159 (3810): 56–6

Johansson-Stenman, O., Svedsater, H., 2008. Measuring Hypothetical Bias in Choice

Experiments: The Importance of Cognitive Consistency. The B.E. Journal of Economic

Analysis & Policy, 8(1), Article 41.

Pelligra, V., Stanca, L., 2013. To Give or Not To Give? Equity, Efficiency and Altruistic

Behavior in an Artefactual Field Experiment. Journal of Socio-Economics 46, pp. 1-9.

Penner, L.A., Dovidio, J.F., Piliavin, J.A. and Schroeder, D.A., 2005. Prosocial

behaviour: multilevel perspectives. Annual Review of Psychology 56, pp. 365–92.

Powell, W., Steinberg, R., 2006. The Nonprofit Sector, A Research Handbook. New

Haven, CT: Yale University Press.

Roberts, R., 1984. A positive model of private charity and public transfers. Journal of

Political Economy 92 (1), pp. 136– 148.

24

Salganik, M., Dodds, P., Watts, D., 2006. Experimental Study of Inequality and

Unpredictability in an Artificial Cultural Market. Science 311, pp. 854-856.

Scharf, K., 2014. Private Provision of Public Goods and Information Diffusion In Social

Groups. International Economic Review 55, pp. 1019-1042.

Shang, J., Croson, R., 2009. A field experiment in charitable contribution: The impact of

social information on the voluntary provision of public goods. The Economic Journal 119

(540), pp. 1422-1439.

Sugden, R., 1984. Reciprocity: the supply of public goods through voluntary

contributions.

Economic Journal 94: pp. 772–87.

Vesterlund, L., 2003. The Informational Value of Sequential Fundraising. Journal of

Public Economics 87(3), pp. 627-657.

Vesterlund, L., 2006. Why do People Give? in Steinberg, R. and Powell, W. (eds.), The

Nonprofit Sector, 2nd edition, New Heaven, Yale Press.

Warr, P., 1982. Pareto optimal redistribution and private charity. Journal of Public

Economics 19 (1), pp. 131 – 138.

Weber, J.M., Kopelman, S., Messick, D.M., 2004. A conceptual review of decision

making in social dilemmas: applying a logic of appropriateness. Personality and Social

Psychology Review, 8(3), pp. 281– 307.

Yariv, L., 2005. I’ll see it when I believe it: A simple model of cognitive consistency.

Working paper, Department of Economics, UCLA.

25

Tables and figures

Table 1. Experimental design: treatments and comparisons.

Table 2. Sample structure.

Source: ISTAT.�Notes: columns 2–4 refer to the survey sample, Sardinia (survey population) and Italy, respectively.

Treatment Treatment Variable Comparisons

(Effect)

NoInfoT No-information Control Group

InfoT Information InfoT vs NoInfoT

(Information effect)

BeliefT No-information

+ Beliefs Elicitation

BeliefT vs NoInfoT

(Belief effect)

Belief&ThresholdT

No-information

+ Beliefs Elicitation

+ Threshold

Belief&ThresholdT

vs BeliefT

(Threshold effect)

Survey Sample Survey Population Italian Population Age 15-29 23.56% 15.80% 15.60% 30-44 26.90% 22.60% 22.20% 45-59 23.27% 22.60% 21.30% 60 and above 26.26% 26.70% 27.10% Gender Male 46.90% 49.00% 48.60% Female 53.10% 51.00% 51.40%

26

Table 3. Completed questionnaires, gender and choices (by Treatment)

Treatment Subjects

approached Compliers Male Female

No Donation

Donation

NoInfoT 388 388 175 213 38 350 (45%) (55%) (9.7%) (90.2%) InfoT 405 405 180 225 15 390 (44.5%) (55.5%) (3.7%) (96.3%) BeliefT 346 250 118 131 7 243 (47.4%) (52.6%) (2.8%) (97.2%) Belief&ThresholdT 266 148 82 66 13 135

(55.4%) (44.6%) (8.8%) (91.2%) 1191

27

Table 4. Descriptive statistics Variable Definition Obs Mean Std, Min Max

Emergency Percent of

donors selecting the organization

in the tax donation choice

1191 0.144 0.351 0 1

Unicef 1191 0.107 0.309 0 1

LAV 1191 0.049 0.215 0 1

WWF 1191 0.037 0.188 0 1

Greenpeace 1191 0.023 0.152 0 1

Amnesty 1191 0.044 0.204 0 1

Caritas 1191 0.236 0.425 0 1

Banco 1191 0.082 0.275 0 1

ADMO 1191 0.217 0.412 0 1

Non-donors 1191 0.061 0.239 0 1

Emergency(rank) Expected rank in terms of

aggregate donations previous year

398 4.007 2.820 1 9

Unicef(rank) 398 3.907 2.516 1 9

Lav(rank) 398 6.439 2.184 1 9

Wwf(rank) 398 5.608 2.158 1 9

Greenpeace(rank) 398 5.638 2.243 1 9

Amnesty(rank) 398 4.927 2.131 1 9

Caritas(rank) 398 3.716 2.498 1 9

Banco(rank) 398 5.389 2.446 1 9

ADMO(rank) 398 5.319 2.755 1 9

demographics

Gender Female=1 1191 0.531 0.499 0 1

Age 1191 44.569 17.703 14 93

Education: middle school 1188 0.27 -share 0 1

Education: high school 1188 0.64 -share 0 1

Education: college level 1188 0.08 -share 0 1 Marital status

Married 1191 0.436 0.496 0 1

Divorced/Separated 1191 0.07 0.250 0 1

Widowed 1191 0.06 0.227 0 1

Ncomp Number of household

members 1191 2.206 1.607 1 11

28

Table 5. OLS Linear-Probability-Model estimation

(1) (2)

VARIABLES Donation Donation

InfoT 0.0699*** 0.0750***

(0.0193) (0.0192)

BeliefT 0.0609*** 0.0601***

(0.0169) (0.0169)

Belief&ThresholdT 0.0101 0.0175

(0.0230) (0.0230)

Age_cent 0.000340

(0.000446)

Edu: High School 0.0586***

(0.0161)

Edu: College 0.0546**

(0.0257)

Female 0.0186

(0.0138)

Married 0.0466***

(0.0156)

Constant 0.902*** 0.849***

(0.0121) (0.0167)

Observations 1,191 1,188

R-squared 0.017 0.039

Dependent variable: Donation (dummy: 1 = donation, 0 = no donation).

Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Age_cent=IndividualAge-AverageAge

29

Table 6. Donations’ distribution (by information)

Table 7. Donations’ distribution clustered (by information)

Cluster A Cluster B Cluster C NoInfoT 25.1% 22.6% 52.3% (88) (79) (183) (#350)

InfoT 20% 11% 69% (78) (43) (269) (#390)

Em

erge

ncy

Uni

cef

LA

V

WW

F

Gre

enpe

ace

Am

nest

y

Car

itas

Ban

co A

lim.

AD

MO

NoInfoT 12.3% 12.9% 6.6% 7.1% 3.1% 5.7% 23.7% 13.7% 14.9%

(43) (45) (23) (25) (11) (20) (83) (48) (52) (#350)

InfoT 13.1% 6.9% 4.1% 0.8% 1.8% 4.4% 27.7% 6.9% 34.4%

(51) (27) (16) (3) (7) (17) (108) (27) (134) (#390)

30

Table 8. Donations’ distribution (by Belief elicitation)

Table 9. Donations’ distribution clustered (by Belief elicitation)

Em

erge

ncy

Uni

cef

LA

V

WW

F

Gre

enpe

ace

Am

nest

y

Car

itas

Ban

co A

lim.

AD

MO

NoInfoT 12.3% 12.9% 6.6% 7.1% 3.1% 5.7% 23.7% 13.7% 14.9%

(43) (45) (23) (25) (11) (20) (83) (48) (52) (#350)

BeliefsT 20.2% 15.2% 4.5% 3.7% 2.9% 3.7% 20.6% 6.6% 22.6%

(49) (37) (11) (9) (7) (9) (50) (16) (55) (#243)

Cluster A Cluster B Cluster C

NoInfoT 25.1% 22.6% 52.3%

(88) (79) (183) (#350)

BeliefsT 35.4% 14.8% 49.8%

(89) (36) (121) (#243)

31

Table 10. “Beliefs – Donation” correlation

Spearman

correlation p-values

Organization_donation – rank1 +0.39 <0.001

Organization_donation – rank2 +0.09 =0.099

Organization_donation – rank3 +0.05 >0.3

Organization_donation – rank4 -0.01 >0.8

Organization_donation – rank5 -0.1 =0.062


Organization_donation – rank7 -0.06 >0.2


Organization_donation – rank9 -0.18 <0.001

32

Table 11. Donations’ distribution (by Threshold)

Table 12. Donations’ distribution clustered (by Threshold)

Em

erge

ncy

Uni

cef

LA

V

WW

F

Gre

enpe

ace

Am

nest

y

Car

itas

Ban

co A

lim.

AD

MO

BeliefsT 20.2% 15.2% 4.5% 3.7% 2.9% 3.7% 20.6% 6.6% 22.6%

(49) (37) (11) (9) (7) (9) (50) (16) (55) (#243)

BeliefsT& ThresholdT

20.7% 14.1% 5.9% 5.2% 2.2% 4.4% 29.6% 5.2% 12.6%

(28) (19 (8) (7) (3) (6) (40) (7) (17) (#135)

Cluster A Cluster B Cluster C

BeliefsT 35.4% 14.8% 49.8%

(89) (36) (121) (#243)

Beliefs&ThesholdT 35.4% 17.8% 47.4%

(47) (24) (64) (#135)

33

Table 13. Bonferroni–Holm stepwise multiple-hypotheses-testing procedure (Holm 1979)

criterion cut-off p-values

rank p-value tag: test level 10% level 5% level 1%

1 0.7 [f] 0.1 0.05 0.01

2 0.3 [e] 0.05 0.025 0.005

3 0.008 [d] 0.033 0.017 0.003

4 0.003 [c] 0.025 0.013 0.003

5 <0.001 [a] 0.02 0.01 0.002

6 <0.001 [b] 0.017 0.008 0.002

34

Table 14. Donations received by each of the nine organizations in the previous fiscal year through the 5X1000 mechanism

Charity Donations received Cluster

EMERGENCY 11,023,415.00 A

UNICEF – ITALIA 5,460,307.00 A

L.A.V. _ LEGA ANTIVIVISEZIONE 1,176,578.00 B

WWF _ WORLD WIDE FOUNDATION ITALIA

1,021,070.00 B

GREENPEACE 758,835.00 B

AMNESTY INTERNATIONAL - SEZIONE ITALIANA

753,674.00 B

CARITAS ITALIANA 193,890.00 C

FONDAZIONE BANCO ALIMENTARE ONLUS

170,351.00 C

ADMO _ ASSOCIAZIONE DONATORI MIDOLLO OSSEO

68,828.00 C

35

Figure 1. Donations’ distribution (by Charity)

0.1

.2.3

.4.5

.6

Emer

genc

y

Unicef

LAV

WW

F

Green

peac

e

Amne

sty

Carita

s

Banco

Alim

enta

re

ADMO

Den

sity

Empirical distribution of the donations

36

Figure 2. Donations’ distribution (by Information)

Figure 3. Donations’ distribution clustered (by Information)

0.1

.2.3

.4.5

.6

Emer

genc

y

Unicef

LAV

WW

F

Green

peac

e

Amne

sty

Carita

s

Banco

Alim

enta

re

ADMO

Emer

genc

y

Unicef

LAV

WW

F

Green

peac

e

Amne

sty

Carita

s

Banco

Alim

enta

re

ADMO

NoInfoT InfoTD

ensi

ty0

.2.4

.6.8

A CB A B C

NoInfoT InfoT

Den

sity

Charities, by cluster

37

Figure 4. Donations’ distribution (by Belief elicitation)

Figure 5. Donations’ distribution clustered (by Belief elicitation)

0.1

.2.3

.4.5

.6

Emer

genc

y

Unicef

LAV

WW

F

Green

peac

e

Amne

sty

Carita

s

Banco

Alim

enta

re

ADMO

Emer

genc

y

Unicef

LAV

WW

F

Green

peac

e

Amne

sty

Carita

s

Banco

Alim

enta

re

ADMO

NoInfoT BeliefsTD

ensi

ty0

.2.4

.6

A CB A B C

NoInfoT BeliefsT

Den

sity


38

Figure 6. Donations’ distribution (by Threshold)

Figure 7. Donations’ distribution clustered (by Threshold)

0.1

.2.3

.4.5

.6

Emer

genc

y

Unicef

LAV

WW

F

Green

peac

e

Amne

sty

Carita

s

Banco

Alim

enta

re

ADMO

Emer

genc

y

Unicef

LAV

WW

F

Green

peac

e

Amne

sty

Carita

s

Banco

Alim

enta

re

ADMO

BeliefT Belief&ThresholdTD

ensi

ty0

.1.2

.3.4

.5.6

A CB A B C

BeliefsT Beliefs&ThresholdT

Den

sity


39

Appendix A: Instructions.

General instructions

Sardegna Solidale and the University of Cagliari are grateful for your decision to participate to

this research. We ask you to fill the following questionnaire in all its parts. Try to answer

autonomously to all questions and, in case of necessity, ask for the support of our researcher.

Answers will be evaluated by us in anonymous form and elaborated in aggregate. Researchers

will not be able in any case to retrieve the respondent’s identity.

The choice questionnaire task

By filling this questionnaire, you will take part in a lottery organized by the Department of

Economics and Business of the University of Cagliari. The drawn will take place on the 31st of

October 2013, one all the questionnaires will be collected. The winner’s identity will be kept

anonymous to the researchers and she/he will be contacted by the Sardegna Solidale volunteers

by means of the numeric code you received.

The first and unique prize is equal to 1000 euros. However, this money will not go to the winner.

She/he can decide whether to give it or not to one charitable organization among those from the

list below. If the prize is not given the winner will not receive anything anyway.

Now we ask you to imagine that you have already won the prize. You now have 1,000 euros

that you could give or not to give to one among the not-for-profit organization listed below.

If after the drawn you will result as the lottery winner, the decision you are about to make will

be implemented for real. That means that if you decided to give to some organization the prize,

such an organization will receive the money for real, otherwise, if you decided not to give, the

prize will not be distributed

40

(in the no-info treatment)

Tick � the box corresponding to your choice:

[Tab. A.1]

Tick your preferred option Organizations

☐ NO DONATION

☐ EMERGENCY

☐ UNICEF – ITALIA

☐ L.A.V. _ LEGA ANTIVIVISEZIONE

☐ WWF _ WORLD WIDE FOUNDATION ITALIA

☐ GREENPEACE

☐ AMNESTY INTERNATIONAL - SEZIONE ITALIANA

☐ CARITAS ITALIANA

☐ FONDAZIONE BANCO ALIMENTARE ONLUS

☐ ADMO _ ASSOCIAZIONE DONATORI MIDOLLO

OSSEO

41

(in the info treatment)


[Tab A.2]


Funding Received in 2011 through 5X1000

(Euros)

☐ NO DONATION

☐ EMERGENCY 11,023,415.00

☐ UNICEF – ITALIA 5,460,307.00

☐ L.A.V. _ LEGA ANTIVIVISEZIONE 1,176,578.00

☐ WWF _ WORLD WIDE FOUNDATION ITALIA 1,021,070.00

☐ GREENPEACE 758,835.00

☐ AMNESTY INTERNATIONAL - SEZIONE ITALIANA 753,674.00

☐ CARITAS ITALIANA 193,890.00

☐ FONDAZIONE BANCO ALIMENTARE ONLUS 170,351.00


OSSEO 68,828.00

42

(in the no-info+beliefs elicitation treatment)

Before making your choice we ask you to order (by assigning a specific rank) each organization

in terms of how much funding you think they received the previous year through the 5X1000

mechanism (denote with 1 the organization that raised more money and with 9 the one that

raised less and with all the other numbers 2-8 those in the intermediate positions)


[Tab. A.3]


Ranking Indicate the position of each organization in term of funding received latest year.

(1 = first… 9 = last)

☐ NO DONATION

☐ EMERGENCY




☐ GREENPEACE





OSSEO

43

(in the no-info+beliefs elicitation + threshold treatment)

Before making your choice we ask you to order (by assigning a specific rank) each organization

in terms of how much funding you think they received the previous year through the 5X1000

mechanism (denote with 1 the organization that raised more money and with 9 the one that

raised less and with all the other numbers 2-8 those in the intermediate positions)

Note that if the amount of the aggregate donations is greater than a given threshold, only a

fraction of the 1000 euros will be actually distributed to the organization.


[Tab A.4]


Ranking Indicate the position of each organization in term of funding received latest year.

(1 = first… 9 = last)

☐ NO DONATION

☐ EMERGENCY




☐ GREENPEACE





OSSEO

diin papr riftp.iza.org/dp10710.pdf · 2017-04-25 · ara iza dp no. 10710 april 2017 information,...

Documents